Forecasting bitcoin price with Greykite

A low-effort forecasting attempt to explore Greykite functionalities

1. Get the data

The Greykite basic data loader

The Greykite basic data plotting -- Interactive!!

2. Baseline model with out-of-box Silverkite

The Greykite basic model config & plotting

Components

Built in backtest, with a long list of metrics!

"run_forecast_config" has automatically performed Train-test-split. The test set have the same length as the forecast horizon, i.e. 30 days

Cross Validation

By default, run_forecast_config provides historical evaluation, so you can see how the forecast performs on past data. This is stored in grid_search (cross-validation splits)

3. Seasonality

weak yearly seasonality

The ability to plot YoY (& other frequency) overlays

4. Change point detection

4.1 Trend change point

4.2 Seasonality change point

Plot all change points together

5. A model with change points & Elon's tweets

Model with Model Components defined

6. A model with gradient boosting

Put together